Isda defends trading book models in response to Basel proposals
The International Swaps and Derivatives Association has proposed an alternative approach to trading book capital requirements, and urged regulators not to disincentivise banks from improving their internal models, in its response to the Basel Committee on Banking Supervision’s review of existing trading book capital rules.
The comment letter has not yet been published by Isda or the Basel Committee, but an industry source provided Risk with a copy.
The 36-page Isda document – which can be accessed via a link at the end of this article – argues regulation should encourage the use of internal models, and that existing proposals do the opposite. It goes on to outline the industry group’s position on several aspects of the review’s proposals – and offer counter-proposals of its own.
“[I]t is important to recognise that firms’ internal models have been significantly upgraded since the financial crisis… Firms should be encouraged to continually develop and improve models, and work with regulators to strengthen modelling standards. This contrasts with certain aspects of the Basel III rules where there appear to be disincentives to modelling,” the Isda comment says.
The committee’s review had proposed desk-level approval of internal models, with the capital charges – based on the expected shortfall metric rather than the current value-at-risk approach – being aggregated according to regulator-set correlation parameters, and possibly subject to a minimum capital floor based on the standardised approach. Banks have argued this would expose them to a cliff effect – soaring capital requirements if individual desks lose approval to model – while the correlation parameters would unfairly restrict diversification benefits.
The Isda response tackles these problems by proposing a smoothly parameterised capital add-on to internal model numbers, intended to prevent sudden capital jumps and avoid overly generous diversification benefit. The industry group claims its alternative would also ensure less divergence from capital numbers generated by the standardised approach.
Under this approach, banks would continue to model their own correlations independently but would add a proportion – known as alpha in the document – of the resulting diversification benefit, defined as the difference between the bank’s full internal model capital and the extreme case of unit correlation between all asset classes. Similarly, at the desk level, a proportion – beta – of the divergence from the standardised charge would be added. Alpha and beta would be set by regulators on a case-by-case basis, based on their confidence in the firm’s models, and could be used as a counter-cyclical measure to lighten capital requirements in difficult times while being more conservative in good times.
“The regulators would be relieved of the burden of setting these correlations – and the responsibility if they don’t work – while giving a conservative limit on how much the model can benefit you through diversification or divergence from the standardised charge. And if need be it can be used to mitigate systemic risk,” says a source familiar with the document.
Isda also argues that flooring trading book capital at a percentage of the standardised approach would increase systemic risk.
“We believe the introduction of regulatory capital floors based on the standardised approach would be harmful to systemic stability, as it risks disincentivising the use and development of internal models. If the standardised approach resulted in a capital floor that would be significantly higher than the floor as calculated under the internal models method, the standardised approach would become the binding constraint. The potential unwanted outcome would be to reduce incentives to improve banks’ internal models, which could not be good for risk management,” the document says.
However, the industry body accepts the review’s recommendation to start using expected shortfall as the new standard risk metric – albeit at a reduced confidence level of 95%, rather than the committee’s proposed 99% – on the basis that it satisfies a mathematical property known as subadditivity, which VAR does not.
Isda and the Basel Committee were not able to respond to requests for comment by press time.
This article was first published on Risk